// the find
yorak/VeRyPy
A python library with implementations of 15 classical heuristics for the capacitated vehicle routing problem.
VeRyPy is a Python library collecting 15+ classical CVRP construction heuristics — Clarke-Wright savings, sweep, nearest neighbor, insertion variants, and others — all faithful to their original 1960s-90s papers. It's aimed at researchers who need a reproducible baseline or a source of initial solutions for metaheuristics, not practitioners who need production-grade routing.
The replication focus is genuine: 11 of 15 algorithms reproduce published numerical results, and the test suite includes 454 benchmark instances to verify this. The code is intentionally loosely coupled — you can import a single algorithm file without dragging in a framework, which is rare in OR libraries. It includes what is likely the only open-source implementation of inter-route 3-opt* (3-opt-star), which is non-trivial. TSPLIB format support means you can benchmark against decades of published problem instances without any data wrangling.
No PyPI package as of 2023 — you install directly from the GitHub zip, which is awkward in any dependency-managed project and breaks reproducibility if the repo changes. Several useful algorithms (GAP, Petal, matching-based savings) require a Gurobi license, which is not free; this is buried in the docs and will catch people off guard. Only symmetric distance matrices are supported, which rules out most real logistics problems with directed road networks or turn penalties. The project has been mostly inactive since late 2023 and the README has multiple TODO blocks still in it, suggesting it was never quite finished.